brynjolfsson, erik & brian kahin (eds.) (2000). understanding the digital economy - data, tools, and...

5/25/2018 Brynjolfsson, Erik & Brian Kahin (Eds.) (2000). Understanding the Digital Economy - Data, Tools, And Research. Cambridge, MA, MIT Press -

1/408


2/408

Understanding the Digital Economy


3/408

This Page Intentionally Left Blank


4/408

Understanding the Digital Economy

Data, Tools, and Research

edited by Erik Brynjolfsson and Brian Kahin

The MIT Press, Cambridge, Massachusetts, and London, England


5/408

2000 Massachusetts Institute of Technology

All rights reserved. No part of this book may be reproduced in any form by anyelectronic or mechanical means (including photocopying, recording, or informationstorage and retrieval) without permission in writing from the publisher.

This book was printed and bound in the United States of America.

Library of Congress Cataloging-in-Publication Data

Understanding the digital economy : data, tools, and research / edited by ErikBrynjolfsson and Brian Kahin.

p. cm.Includes bibliographical references and index.ISBN 0-262-02474-8 (hc : alk. paper)1. Electronic commerceCongresses. I. Brynjolfsson, Erik. II. Kahin, Brian.

HF5548.32 .U53 2000330.9dc21 00-033947


6/408

Contents

Introduction 1Erik Brynjolfsson and Brian Kahin

The Macroeconomic Perspective

Measuring the Digital Economy 13

John Haltiwanger and Ron S. JarminGDP and the Digital Economy: Keeping up with theChanges 34Brent R. Moulton

Understanding Digital Technologys Evolution andthe Path of Measured Productivity Growth: Present

and Future in the Mirror of the Past 49Paul A. David

Market Structure, Competition, and the Role ofSmall Business

Understanding Digital Markets: Review and

Assessment 99Michael D. Smith, Joseph Bailey, and Erik Brynjolfsson

Market Structure in the Network Age 137Hal R. Varian

The Evolving Structure of Commercial InternetMarkets 151

Shane GreensteinSmall Companies in the Digital Economy 185Sulin Ba, Andrew B. Whinston, and Han Zhang


7/408

vi

Contents

Small Business, Innovation, and Public Policy in

the Information Technology Industry 201Josh Lerner

Employment, Workforce, and Access

Technological Change, Computerization, and theWage Structure 217

Lawrence F. KatzThe Growing Digital Divide: Implications for anOpen Research Agenda 245

Donna L. Hoffman and Thomas P. Novak

Extending Access to the Digital Economy to Ruraland Developing Regions 261

Heather E. Hudson

Organizational Change

IT and Organizational Change in DigitalEconomies: A Sociotechnical Approach 295Rob Kling and Roberta Lamb

Organizational Change and the Digital Economy:A Computational Organization Science Perspective 325Kathleen M. Carley

The Truth Is Not Out There: An Enacted View ofthe Digital Economy 352Wanda J. Orlikowski and C. Suzanne Iacono

Contributors 381

Index 389


8/408

Introduction

Erik Brynjolfsson and Brian Kahin

The digital economydefined by the changing characteristics of infor-mation, computing, and communicationsis now the preeminent driver

of economic growth and social change. With a better understanding ofthese fundamental transformations, we can make wiser decisionswhetherwe are investing in research, products, or services, or are adapting ourlaws and policies to the realities of a new age.Neal Lane, Assistant tothe President for Science and Technology, April 1999

Although there is now a substantial body of literature on the roleof information technology in the economy, much of it is inconclu-

sive. The context is now changing as the success of the Internet andelectronic commerce (e-commerce) introduces new issues ofinfluence and measurement. Computers created a platform for thecommercial Internet; the Internet provided the platform for the

Web; the Web, in turn, provided an enabling platform for e-commerce. The Internet and the Web have also enabled profoundchanges in the organization of firms and in processes within firms.

The Internet links information to locations, real and virtual. Itlinks the logic of numbers to the expressive power and authority of

words and images. Internet technology offers new forms for socialand economic enterprise, new versatility for business relationshipsand partnerships, and new scope and efficiency for markets.

The commercial Internet has only had about six years to play outin earnest, but the numbers show a remarkable accelerationadoubling of Internet connections year after year and, more re-cently, a variety of figures on e-commerce showing even fastergrowth. Web transaction costs are as much as 5099 percent less


9/408

2Brynjolfsson and Kahin

than conventional transaction costs.1It is this chain of drivers and

its implications for the economy and society as a whole that leadsus to speak of a digital economy.

The term information economyhas come to mean the broad,long-term trend toward the expansion of information- and knowl-edge-based assets and value relative to the tangible assets andproducts associated with agriculture, mining, and manufacturing.The term digital economyrefers specifically to the recent and still

largely unrealized transformation of all sectors of the economy bythe computer-enabled digitization of information.

Because of its mandate in matters of interstate commerce andforeign trade, the federal government has primary responsibilityfor evaluating the health and direction of the economy. Theemerging digital economy makes commerce less local, more inter-state, and, especially, more global, in line with a long-term trend

toward market liberalization and reduced trade barriers. At thesame time, the picture presented by public information sources isbecoming less and less complete. What we know about e-commercecomes from proprietary sources that use inconsistent methodolo-gies. Economic monitoring, like policy development, is challengedby quickly evolving technologies and market practices.

The nature and scope of the digital economy are matters ofconcern to nations at all levels of development. Like consistentlegal ground rules, an open, testable platform of public economicinformation is essential to investment and business decisions. It isalso essential to sound monetary policy and to setting taxes andspending budgets. Ultimately, understanding the digital economyis relevant to a wide range of policies: R&D investment, intellectualproperty, education, antitrust, government operations, account-ing standards, trade, and so on.

All countries must confront the unfettered flow of informationon the Internet and the ease with which international transactionsand investments can take place. While the digital economy isknown as a generator of new business models and new wealth, it isalso undermining old business models and threatening invest-

ments and jobs in certain established businesses. With the excite-ment comes anxiety and concern about the how the ingredients ofthe digital economy should be configured for optimal advantage.


10/408

Introduction3

Outside the United States, it is sometimes viewed as a suspect

phenomenon, deriving in part from American strengths in com-puter technology and software, flat-rate phone service, and thescale advantages of the English language. For all these reasons, itbegs investigation.

In April 1998, the U.S. Department of Commerce issued TheEmerging Digital Economy, a landmark report that recognized theaccelerating importance of the Internet and e-commerce in the

national economy. Bearing the imprimatur of the federal govern-ment, the report offered new perspective on the role of informa-tion technology in productivity, inflation, economic growth, andcapital investment. It has been cited frequently and succeeded bya number of reports assessing these and other developments.2

In November 1998, as part of the second phase of an initiative onglobal electronic commerce, President Clinton charged the assis-

tant to the president for economic policy to undertake an assess-ment of the digital economy. In addition to asking the Departmentof Commerce to update The Emerging Digital Economy, the presidentasked that experts be convened to assess the implications of thedigital economy and to consider how it might best be measured andevaluated in the future. Accordingly, an interagency workinggroup on the digital economy planned a public conference, whichtook place on May 2526, 1999, at the Department of Commerce(www.digitaleconomy. gov). The conference was sponsored by theDepartment of Commerce, the National Science Foundation, theNational Economic Council, the Office of Science and TechnologyPolicy, and the Electronic Commerce Working Group, the um-brella interagency group for the administrations global e-com-merce initiative.

The conference sought a common baseline for understandingthe digital economy and considered how a clearer and more usefulpicture of that economy might be developed. While recognizingthe convergence of communications, computing, and informa-tion, the conference looked beyond those sectors to focus on thetransformation of business and commerce, processes and transac-

tions, throughout the economy.This books four parts mirror the four basic topics considered atthe conference:


11/408


The macroeconomic perspective:How do we measure and assess the

digital economyand its implications for the economy as a whole? The texture of the digital economy:How do firms compete and howdo markets function, and how is this different from traditionalcompetition? What are the opportunities for and impediments tothe participation of individuals and small businesses?

The impacts on labor demand and participation:Do the new tech-

nologies exacerbate inequality? What skills, technologies, andinstitutions are needed to support broader access to the benefits ofthe digital economy by different individuals and groups?

Organizational change:How does the digital environment affectthe structure and operation of firms and institutions?


Information technology is playing an increasing role in growth,capital investment, and other aspects of the economy. The scopeand significance of these transformations remain open to question,however, in large part because underlying measurement and meth-odology problems have not been resolved.

How should we identify and measure the key drivers of the digitaleconomy?

What are the industry-level and economy-wide investments re-lated to e-commerce, including investments in information tech-nology equipment and workers?

What are the implications for growth, employment, productivity,and inflation?

How should we account for intangible consumer benefits andburdens?

There are three chapters in this part. In Measuring the DigitalEconomy,John Haltiwanger and Ron Jarmin note that the emer-gence of e-commerce is part of a broad spectrum of changes overseveral decades related to advances in information technology and

the growth of the broader digital economy. After reviewing thecurrent activities of federal statistical agencies, they conclude thatcurrent data collection activities are inadequate and provide some


12/408

Introduction5

practical advice on how to improve measurement of the digital

economy.In GDP and the Digital Economy: Keeping up with the Changes,

Brent Moulton argues that inadequate measurement of the trueoutput of the digital economy has contributed to past difficultieseconomists have had in identifying the productivity benefits of theIT revolution. He shows that despite these measurement difficul-ties, the measured contribution of computers to GDP has grown

substantially in the late 1990s, and he outlines an agenda forimproving research in this area.

In a seminal paper a decade ago, Paul David noted that newtechnologies such as electric motors or computers require enor-mous complementary investments, such as changes in organiza-tional structure, in order to reach their full productive potential.3

In his chapter, Understanding Digital Technologys Evolution

and the Path of Measured Productivity Growth: Present and Futurein the Mirror of the Past,David provides a detailed review of thesubsequent literature and shows how much of the micro and macroevidence on IT and productivity affirms the importance of organi-zational complements.

Market Structure, Competition, and the Role of Small Business

The digital economy includes information and communicationstechnology, e-commerce, and digitally delivered services, software,and information. The characteristics of these goods and services(including factors such as economies of scale, network effects,public good characteristics, and transaction costs) can lead todifferent market structures and competitive conditions. Unfortu-nately, such characteristics are difficult to measure, technologiesare changing rapidly, and relevant market boundaries are fluid anddifficult to define. Some have speculated that the Internet and e-commerce hold great promise for small firms, by liberating themfrom proprietary value chains, diminishing transaction costs, andproviding access to global markets, but without adequate data it is

difficult to test this speculation. What are the relationships and interactions between the eco-nomic characteristics of digital technologies, products, and ser-


13/408


vices and the structure and competitiveness of markets?

What are the key determinants of prices (overall price levels,price flexibility, price dispersion, etc.), market structure and effi-ciency (competitive, noncompetitive, segmented, etc.), and com-petition (price based, market share based, etc.)?

What roles do startups and small firms play in different segmentsof the digital economy? What are the barriers to launching and

growing small firms? How and to what extent do the Internet and e-commerce eitherbenefit or handicap entrepreneurs and small- to medium-sized firms?

The five chapters in this part review the empirical evidence onhow competition and strategy differ in the digital economy. Two ofthe chapters specifically look at the changing role of smaller firms.

In Understanding Digital Markets: Review and Assessment,Michael Smith, Joseph Bailey, and Erik Brynjolfsson summarize therecent literature on how the Internet is affecting competition andmarket efficiency. They start with findings for several dimensionsof market efficiency and then focus on the puzzling finding ofunusually high price dispersion on the Internet. They conclude

with a set of developments to watch and provide an annotated

appendix of research on the Internet and competition.In Market Structure in the Network Age,Hal Varian shows howseveral fundamental principles of economics can be used to in-crease understanding of how e-commerce changes competition.He analyzes versioning, loyalty programs, and promotions, in eachcase illustrating his points with examples from e-commerce andoutlining the research issues raised.

Shane Greenstein admirably demonstrates the value of develop-ing new data sources in his chapter, The Evolving Structure ofCommercial Internet Markets.He focuses on the commercializa-tion of a key link in the e-commerce value chain: the InternetService Providers (ISPs) who supply access to the Internet formillions of consumers and businesses. Using this example, heanalyzes a set of broader questions that are important for research-

ers, policymakers and managers.In Small Companies in the Digital Economy,Sulin Ba, AndrewWhinston, and Han Zhang outline some of the Internets special


14/408

Introduction7

opportunities and challenges for smaller enterprises. They focus

on the way information asymmetries on the Internet enhance theimportance of branding and of trusted third parties, and theydescribe some significant technologies that are likely to help withthese issues.

In Small Business, Innovation, and Public Policy in the Informa-tion Technology Industry,Josh Lerner documents the ambiguousoverall role of small business in innovation but shows that a

particular subset of small businessesfirms that are venturebackedhave been particularly strong innovators. He focuses onthe concentration of venture financing in IT industries and con-cludes by discussing recent changes in intellectual property lawsthat appear to favor larger firms, drawing some implications forpolicy makers.

Employment, Workforce, and Access

As information and communications technologies transform theglobal economy, they are changing the U.S. workforce in terms ofsize, composition, and the knowledge and skills required forsuccess. Indeed, the competitiveness of nations and companiesappears increasingly dependent on the ability to develop, recruit,and retain technologically sophisticated workers. There are con-cerns that the U.S. workforce is already unable to meet the marketdemand for skilled and knowledgeable workers and that this gap isgrowing. Furthermore, there is growing concern that the benefitsof the digital economy are not equitably shared, giving rise to adigital divide. There are a variety of options for overcomingbarriers to participation, and it is important to understand theextent to which such options are available, utilized, and cost-effective.

How reliable are current models for projecting the size andcomposition of labor markets in occupations where technologiesare changing rapidly? How can they be improved?

How does the growth of e-commerce and investment in theInternet and related technologies affect the level and compositionof labor market demand? How can these influences be untangledfrom other factors?


15/408


What can be learned from firm-level or industry-level studies as

compared to aggregate labor market models? What barriers impede the diffusion of e-commerce across thesociety?

To what extent and in what ways does e-commerce enhance,preserve, or diminish diversity? To what extent does e-commerce

work to increase or lessen opportunities for economic progress for

disadvantaged individuals, groups, and regions?The three chapters in this part raise troubling questions about

growing inequality and underscore that the benefits of the digitaleconomy are not necessarily evenly spread among different groupsin society.

In Technological Change, Computerization, and the WageStructure, Larry Katz discusses one of the most troubling eco-

nomic phenomena of the past two decades. Wage inequality hasexpanded dramatically, making the rich even richer relative to thepoor. Katz notes that this widening inequality has coincided withgrowing use of IT and is particularly closely linked to increasedrelative demand for more educated and skilled workers. He reviewsthe existing literature and suggests some new empirical approachesthat might help us identify the relationships among computeriza-

tion, demand for skilled labor, and income inequality.Donna Hoffman and Thomas Novak summarize a range of

statistical evidence in The Growing Digital Divide: Implicationsfor an Open Research Agenda.They highlight the differentiallevels of computer adoption and Internet usage among variousdemographic groups. The provocative facts they review raise im-portant questions for researchers and policy makers who areconcerned about the potential gap between information havesand have-nots.

In Extending Access to the Digital Economy to Rural andDeveloping Regions, Heather Hudson examines opportunitiesfor extending Internet access to disadvantaged groups in industrialnations and also to populations in developing nations. She docu-ments some striking disparities in basic measures of access, such astelephone lines, and provides a useful guide to future research inthis area as well as an appendix summarizing some of the availabletechnological options.


16/408

Introduction9

Organizational Change

While information technology is routinely deployed in organiza-tions to reengineer processes, gain strategic new advantages, ornetwork across boundaries, it may also produce unintended out-comes. With the rise in interorganizational systems, e-commerce,and new organizational forms, questions arise about how newrelationships among suppliers, customers, competitors, and pro-

viders will be crafted and what these new configurations imply forexisting organizations.

How will a digital economy affect structure and relationshipswithin and among firms?

To what extent and under what conditions will a digital economylead to new organizational cultures?

How will a digital economy affect stratification within and acrossfirms?

The three chapters in this part look at the question of IT andorganizational change from three different perspectives. In ITand Organizational Change in Digital Economies: A Sociotechnical

Approach,Rob Kling and Roberta Lamb argue that information

systems require substantial organizational changes before theybecome fully effective. Through a series of insightful case studies,they highlight how this perspective diverges from the alternative

view that treats IT largely as a tool. They call for a program oflongitudinal research on the interaction of IT, organizations, andoutcomes.

Kathleen Carley draws on research from Carnegie Mellon and

elsewhere in her chapter, Organizational Change and the DigitalEconomy: A Computational Organization Science Perspective.She characterizes the emerging intelligence spaces from theperspective of computational organizational science and showshow simulations can help us understand the nature of social andeconomic interactions as commerce becomes electronic, agentsbecome artificial, and more and more of the world becomes digital.

This part and the book conclude with a cautionary perspectivefrom Wanda Orlikowski and Suzanne Iacono. In The Truth Is NotOut There: An Enacted View of the Digital Economy,they stress


17/408


that the digital economy is not an immutable and inevitable object,

subject to dispassionate analysis, but rather an ever-changing socialconstruction. This has important implications for researchers, whoneed to be cognizant of the complex and often nonlinear relation-ships they are studying. It also serves as an essential reminder to usall that we have not just the opportunity but the responsibility toshape the digital economy in ways that reflect our values and goals.

Notes

1. OECD, The Social and Economic Implications of Electronic Commerce(1998), p. 63(Table 2.4).

2. Lynn Margherio et al., The Emerging Digital Economy (Department of Com-merce, April 1998);Fostering Research on the Economic and Social Impacts of Informa-tion Technology(Washington, DC: National Academy Press, 1998); Economic and

Social Significance of Information Technologies,in National Science Founda-tion, 1998 Science And Engineering Indicators; The Economic and Social Impacts ofElectronic Commerce(OECD, September 1998); David Henry et al., The EmergingDigital Economy II (Department of Commerce, June 1999).

3. Computer and Dynamo: The Modern Productivity Paradox in a Not-Too-Distant Mirror,in Technology and Productivity: The Challenge for Economic Policy,Paris: Organization for Economic Co-operation and Development (1991), pp.315348.


18/408



19/408

This Page Intentionally Left Blank


20/408

Measuring the Digital Economy

John Haltiwanger and Ron S. Jarmin

Introduction

This chapter focuses on the data needs and measurement chal-lenges associated with the emerging digital economy. We muststart, however, by defining what we mean by the digital economy.The dramatic growth of what is being called electronic commerce(e-commerce) has been facilitated by the expansion of access tocomputers and the Internet in workplaces, homes, and schools.There is a broad consensus that computers and the Internet are

producing rapid changes in how goods and services are produced,the nature of the goods and services being offered, and the meansby which goods and services are brought to market. We view theemergence of e-commerce, however, as part of a broad spectrum ofchanges in the structure of the economy related to developmentsextending over several decades in information technology (IT).U.S. statistical agencies are still addressing the challenges of mea-suring the changes brought on by the IT revolution. For measure-ment purposes, the challenges brought on by the growth ofe-commerce are closely linked to those brought on by advances inIT.

The banking sector provides a good example of the problemsconfronting statistical agencies. The IT revolution has led to theintroduction of new services such as electronic banking and ATMs.Statistical agencies have struggled with how to define and measureoutput in banking for years, and the IT revolution has done


21/408

14Haltiwanger and Jarmin

nothing to ease the struggle. For example, ATMs allow customers

to access their accounts 24 hours a day 7 days a week while reducingor eliminating the time they spend in teller lines. This clearlyrepresents an increased level of customer service. Yet the value ofsuch services is not directly measured in any official statistics,

whereas the cost of installing ATM networks is. Because of measure-ment problems of this sort, government statistics understate theproductivity increases in banking that come from investments in

IT.There is widespread belief that we need to make significant

changes to the U.S. statistical system in order to track the growthand impact of the digital economy. The 1997 Department ofCommerce report on The Emerging Digital Economy provides ex-amples of aspects of the digital economy that we should be measur-ing:

1. The shape and size of the key components of the evolving digitaleconomy, such as e-commerce and, more generally, the introduc-tion of computers and related technology in the workplace.

2. The process by which firms develop and apply advances in IT ande-commerce.

3. Changes in the structure and functioning of markets, includingchanges in the distribution of goods and services and changes inthe nature of international and domestic competition.

4. The social and economic implications of the IT revolution, suchas the effects of IT investments on productivity.

5. Demographic characteristics of user populations.

After presenting what we believe are the data needs for assess-ments of the digital economy, we will summarize the currentactivities of federal statistical agencies. Not surprisingly, we willargue that current data collection activities are inadequate and thata number of difficult issues need to be resolved to improve thesituation. We will offer some practical and feasible examples of

what statistical agencies can do to improve measurement, but we

stop short of providing specific suggestions and instead describe aframework in which discussions about changes to the measure-ment system can take place. This process needs to begin soon


22/408

Measuring the Digital Economy15

because of the considerable lag that often occurs between identify-

ing a data need, finding a way to address it, implementing acollection program, and getting data to users.

Data Needs for the Information Economy

We will restrict our attention to the types of data that are requiredfor public policy and general economic research and that are

typically collected by government statistical agencies through na-tionally representative surveys of individual, household, and busi-ness units. We recognize that there is a large data-using constituencythat requires types of data different from those collected by thestatistical agencies. This constituency has traditionally been servedby private-sector sources, and we believe that this will continue tobe the case.

Given the pace of change in IT and the myriad new ways in whichbusinesses, households, and others exploit IT, it is understandablethat the institutions that collect economic and demographic dataare behind in measuring the magnitude and scope of ITs impacton the economy. But before discussing measurement issues di-rectly related to IT and the digital economy, we need to stress thatimproved measurement of many traditional items is crucial if weare to understand fully ITs impact. It is only by relating changes inthe quality and use of IT to changes in traditional measures such asproductivity and wages that we can assess ITs impact on theeconomy. For example, if we cannot measure and value output inthe service-sector industries where IT is important, it will bedifficult to say anything about its impact. Thus, as part of theattempt to improve measurement of the digital economy, we alsoneed better ways to measure the activities of firms in the so-calledunmeasured sectors of the economy (e.g., services) and to improvethe quality of statistics for the measured (i.e., the goods-producing)sectors.

Three broad areas of research and policy interest related to thedigital economy require high-quality data. First, there is the inves-

tigation of the impact of IT on key indicators of aggregate activity,such as productivity and living standards. Aggregate productivitygrowth slowed over much of the period in which large investments


23/408


in IT occurred, especially in service industries, such as banking,

that had particularly large IT investments. A number of studies, atvarious levels of aggregation, failed to find a link between ITinvestments and productivity, leading to the identification of aproductivity paradox(Solow 1987; Berndt and Morrison 1995;for a review of the literature on the link between IT investments andproductivity see Brynolfsson and Yang 1996).

Several explanations have been offered for this paradox. One is

that official statistics do not capture all the changes in output,quality, and cost savings associated with IT and therefore under-state its impact (Siegel and Griliches 1994). Another compares ITto previous innovations in the economy, such as electrification, andnotes that there can be a considerable lag between investments insuch innovations and related productivity increases (David 1990;Greenwood and Yorgulu 1997).

Recent studies using data from a variety of sources have in factreported a link between IT and productivity (e.g., Jorgenson andStiroh 1995; Greenan and Mairesse 1996; Brynjolfsson and Hitt1995, 1996; Dunne et al. 1999). These, combined with improvedaggregate productivity performance, have led some to speculatethat productivity is no longer a paradox (Anonymous 1999). Whileit is undoubtedly the case that several firms and industries havefinally seen returns on investments in IT, the empirical evidencefor an economy-wide impact is limited. A large part of this limita-tion, though, may be due to the inadequacy of available data.

With the growth of e-commerce, particularly in business-to-business transactions, we are no longer interested only in measur-ing the impact of computers and IT on productivity withinorganizations. We now want to assess whether there have beenmeasurable increases in productivity related to improvements ininformation flows and reduced transaction costs between organiza-tions that do business electronically. We want to see whether e-commerce is associated with measurable productivity gains insectors and firms that rely heavily on e-commerce with respect tothose that employ e-commerce less extensively.

Of related interest are the implications of IT and e-commerce forthe measurement of the capital stockparticularly equipment.For accuracy we need measurements of equipment investment by


24/408


detailed asset category, quality-adjusted deflators for such invest-

ment that take into account advances in technology, and appropri-ate measures of the depreciation rates of the assets in question. Inthe case of IT, the measurement of depreciation rates has becomemuch more difficult due to the rapid pace of the changes involved(e.g., the rate at which the speeds of successive generations ofprocessors increase) and the associated rapid turnover of com-puter hardware and software. Storage closets, attics, and junkyards

are increasingly cluttered with PCs that were on the cutting edgejust a few years ago! While it is important to measure the nationalcapital stock, we must also understand wherein what industries,geographic locations, and types of firmsIT is being applied. This

will provide a basis for evaluating the impact of IT on productivitybecause, in principle, we should observe the greatest gains inproductivity in those sectors that apply IT most effectively. This

suggests that using accounting methods to estimate IT (or othertypes of) investment is insufficient, since these analyses requiremicro-level data. For this reason, data on IT investment must becollected from businesses and other organizations in every majorsector of the economy.

The second area of research and policy interest that requireshigh-quality data is the impact of IT on labor markets and incomedistribution (for broader discussions of these issues see OECD 1999and DOC 1999). Of particular interest here is the issue of whetherIT is increasing wage and income dispersion by creating groups ofhaves and have-nots based on whether people have the skills and/or are employed in the appropriate sectors to take advantage of ITadvances (Autor, Katz, and Krueger 1997; Dunne et al. 1999).

Answering this question requires measuring the use of computersand other IT equipment in the workplace and relating it to wages.It would also be useful to assess whether or not the educationalsystem is providing the next generation of workers with the skillsneeded to succeed in the digital economy.

Third, many people would like to assess the impact of IT on theway production is organized. They want to understand how firm

and industry structures have changed as IT has become a moreimportant input to production in every sector of the economy (Hittand Byrnolfsson 1997). And, most importantly, they want to under-


25/408


stand the impact of the digital economy on market structure. There

is a growing sense that e-commerce is dramatically changing theways in which buyers and sellers find and interact with each other.Electronic networks in the form of Electronic Data Interchanges(EDIs) have existed for some time, allowing companies to commu-nicate with major suppliers and customers. Until recently, how-ever, EDIs were limited primarily to large firms with mainframecomputers that communicated across expensive proprietary lines.

The Internet allows anyone with a PC and modem to communicatewith millions of computers worldwide. This has important implica-tions for the nature and location of businessesparticularly thoseinvolved in the distribution of goods and servicesand for howmarkets work.

The availability of inexpensive yet powerful computer hardwareand software reduces the costs of setting up an e-business and

expands the possibilities for siting businesses. The open structureof the Internet now allows small firms to download specificationsand bid on jobs previously available only to a select few who hadaccess to EDIs. This is likely to have significant market structureimplications for a wide array of goods and services.

At the same time, the Internet is giving consumers more powerin the marketplace by making information on the prices andqualities of a wide range of goods and services more accessible.Price competition could be substantially enhanced when buyerscan easily search for alternative suppliers of goods and services.

It is also important to get a handle on the degree of substitutionoccurring between goods and services purchased through e-com-merce (e.g., from Amazon.com) and similar goods and servicespurchased through traditional channels (e.g., from a neighbor-hood bookstore). This substitution may be particularly importantfor digitalgoods and services. Digital goods, which will eventuallyinclude books, movies, and music, are goods that can be deliveredto customers in digital form over the Internet. Such goods cantheoretically bypass traditional distribution channels. This obvi-ously has major implications for the wholesalers, retailers, and

transporters of this class of products. Researchers will want to keeptrack of changes in how these products are delivered as thebandwidth of the Internet expands.


26/408


We can now summarize the general data requirements for the

digital economy. We need statistics on inputs and outputs that willallow us to construct measures of productivity at several levels ofaggregation, to maintain the National Income and Product Ac-counts, to conduct cross-region and industry studies, and to per-form micro-level data analyses. This includes the construction ofappropriate quality-adjusted price deflators. We are interested inunderstanding not only the implications for consumer and pro-

ducer prices but also whether market competition (as reflected, forexample, in price-cost markups) has changed as a result of e-commerce. We also need to understand the organization andlocation of production and where workers of different types work.This requires collecting at least some data at the subfirm, orestablishment, level. We also need data on the human capitalembodied in workers and on the occupations and industries they

work in and the wages they receive. Finally, we need detaileddemographic data on the U.S. population, and in particular onindividuals and households that participate in the digital economy.

Assuming that we will continue to collect and improve ourtraditional menu of economic and demographic data, and giventhe three broad research areas in which we would like to assess theimpact of IT, what are some of the specific data items we should bemeasuring in order to keep track of the digital economy? Webelieve that there are five areas where good data are needed. Theseare: (1) measures of the IT infrastructure, (2) measures of e-commerce, (3) measures of firm and industry organization, (4)demographic and labor market characteristics of individuals usingIT, and (5) price behavior. Boxes 15 give examples of specific dataitems of interest to policymakers and researchers in each of thesefive areas.

How Well Are We Measuring the Digital Economy?

Because we cannot survey all data sources, we will focus on datacollected by the Census Bureau and other federal statistical agen-

cies. (In several cases, data relevant to the digital economy areavailable from sources outside the federal statistical system. Thesedata sets tend to be specialized, are often based on nonrepresenta-


27/408


Box 1

Data Needs for the Digital Economy: Information TechnologyInfrastructure

We should measure the physical and software infrastructure of theinformation economy. In particular, data collection efforts should focus oninvestments in physical infrastructure (e.g., IT equipment includingcomputers, phone lines, switches, fiber optic and cable lines, satellites,wireless networks, and LAN equipment). We should also measure invest-ments in software infrastructure. We should collect data on the capacity of

the Internet and other networks as well as the actual traffic on thesesystems. It is crucial that we measure depreciation in infrastructure (bothequipment and software) and how investments and depreciation act tochange the capacity of the digital infrastructure. And we need to havesome idea of the IT and software components of non-ITequipment suchas numerically controlled machines.

tive surveys, and are rarely available to the wider research and

policy communities.) Even though our survey is incomplete, itshould be apparent that current data collection for the itemsoutlined in the last section is spotty and inconsistent.

Infrastructure

Our estimates of the impact of computers and related information

technologies are based on relatively limited data sets. As with mostequipment investment, we measure the magnitude of aggregateinvestment in computers by examining the output of sectorsproducing such equipment and adjusting for exports and imports(i.e., the statistics are generated from Current Industrial Reportsand export and import statistics, as well as annual surveys ofbusinesses). This accounting methodology provides reasonable

national totals of investment in computers and related technolo-gies on a nominal basis. Much work has been done to generatequality-adjusted deflators for computers, and to the extent thatthese deflators are reliable, a reasonable estimate of the nationalreal investment in computers emerges. However, we know verylittle about what types of firms and industries are implementingcomputers and other advanced technologies. In the past, the

Annual Survey of Manufactures (ASM) asked about computerinvestment in economic census years (in 1977, 1982, 1987, and


28/408


1992). This question was not asked in 1997 but will probably beasked again in the future. The ASM also asks about purchasedcommunication services and software purchases. Every five years,as part of the economic census, the Census Bureau conducts theBusiness Expenditure Survey (formerly known as the Assets andExpenditures Survey) for businesses in Retail, Services, and Whole-sale. This survey contains a question about spending on computers,peripherals, and software. For multiunit companies, the unit ofanalysis in this survey is not necessarily either the firm or theestablishment. Rather, data are collected roughly at the level of alegal entity (as defined by Employer Identification Numbers) orline of business. (An example would be the drugstore operations ofa company that operates in several retail markets.)

In the past, the Annual Capital Expenditure Survey (ACES) didnot break out equipment investment by type of capital, but it will

soon begin to do so. Because this survey is at the firm level and manylarge, multiunit firms span several industries and regions, it will bedifficult to use the results to construct accurate statistics for invest-ments in IT and other types of capital by industry and geographicregion. (The 1998 survey asked companies to break out equipmentby both type of equipment and industryroughly at a 2-digit level.)The Bureau of Economic Analysis (BEA) produces data on capital

expenditures and stocks by asset type by industry. However, theallocation of assets by industry are derived from Capital Flowallocation tables that are based on strong assumptions and limited

Box 2

Data Needs for the Digital Economy: E-Commerce

We should measure e-commerce by the magnitude and type of bothbusiness-to-business and business-to-consumer electronic transactions. Weshould also try to measure separately digital and nondigital goods andservices. Nondigital products must be physically delivered to consumers.Digital products can bypass the wholesale, retail, and transport network.Also, digital products may have very different (nonlinear) pricing struc-tures due to their high fixed costs and low marginal costs (Shapiro and

Varian 1999). This may be important for computing valid price deflatorsand may make it difficult to use revenue-based measures of activity levels.We should also measure the use of e-commerce for both transactions andnontransaction purposes (e.g., customer service, general information, andbid posting).


29/408


data (e.g., the asset allocations by industry are based in part on theoccupational mix within industries). In short, while we may have areasonable national estimate of investment in computers, we knowlittle about investment in computers by industry, geographic area,or firm type.

There is little official data on the investments in and the capacityof the telecommunications networks that support the Internet.

There is also little information outside of the ASM about invest-ments in software. It is especially important to get a handle on thedifferential pricing and depreciation of software. Without thisinformation, it will be virtually impossible to get an accuratemeasure of the service flow of software investments.

E-Commerce

There is even less information collected on e-commerce. It isimportant to emphasize that e-commerce sales should be coveredby economic censuses and surveys since the Census Bureau main-tains representative samples of all businesses, including thoseengaged in e-commerce. In the past, however, there has been nosystematic attempt to break out sales by method of selling. Thus, weknow how much firms engaged in e-commerce sell, but not howmuch they sell via e-commerce.

The Census Bureau has begun to inquire about e-commerce saleson many of its monthly and annual business surveys. While there isconsiderable interest in separately measuring business-to-consumerand business-to-business e-commerce transactions, currently noCensus Bureau survey elicits such information. As for digital goodsand services, there is currently no way to estimate the value of salesin which the good or service being transacted is delivered to thepurchaser electronically.

Box 3

Data Needs for the Digital Economy: Firm and Industry StructureWe should measure the impact of improvements in IT, software, and theInternet on firm and market structures. More generally, we should quantifythe changes in the location, industry, size, and organizational structure ofbusinesses, as well as changes in their input mix (e.g., capital, labor,inventories) and their relationships with other businesses (e.g.,outsourcing).


30/408


Firm and Industry Structure

The ingredients for characterizing the changing structure of mar-kets in terms of the location of businesses, the industries in whichbusinesses operate, and the size distribution of businesses are

available in business lists maintained by federal statistical agencies.For example, the Census Bureau maintains the Standard StatisticalEstablishment List (SSEL), which is constructed from administra-tive data, economic censuses, and surveys. The SSEL follows theuniverse of all establishments in the United States and is a veryuseful resource for keeping track of the changing demography (interms of size, location, and industry) of U.S. businesses. It is an

underutilized resource for this type of analysis. For example, thereis some sense that e-commerce has reduced entry barriers substan-tially, allowing small businesses to compete in an unprecedentedmanner. Because the SSEL offers a comprehensive dynamic pic-ture of all businesses (large and small), it is a superb resource fortracking the impact of the digital economy on small businesses.There is also an ongoing collaborative project between the SmallBusiness Administration and the Census Bureau to develop and usea longitudinal version of the SSEL to track the dynamics of small vs.large businesses.

There are some challenges in the use of the SSEL for these typesof analyses. First, the quality of the analyses depends critically onthe quality of the industry and location codes in the SSEL. Whilethe quality of such codes is relatively high for most businesses, thequality for new and small businesses is lower. This could prove to beproblematic for tracking the impact of the digital economy becauseof its dynamic nature and purportedly large number of small start-

Box 4

Data Needs for the Digital Economy: Demographic and WorkerCharacteristics

We should measure the demographic and labor market characteristics ofindividuals and workers and compare those participating in the digitaleconomy to those not participating. In particular, we should measurecomputer use at school, work, and the home and relate these to measuresof economic outcomes such as wages and assets and to demographiccharacteristics such as education, occupation, gender, race, age, and placeof residence.


31/408


ups. In addition, while the new North American Industrial Classi-fication System (NAICS) offers much greater detail in terms ofindustries in the information and service sectors, it is unclear howeasy it will be to track key aspects of the digital economy withoutadditional modifications to our industry codes. For example, thereare no current plans to classify businesses that primarily sell by e-

commerce in a separate category. Instead they are grouped withmail-order houses.

Demographic and Worker Characteristics

The Current Population Survey (October supplement every threeyears) looks at household computer use. This information has

enabled analysis of the impact of computer use on labor marketoutcomes, such as wages, and better understanding of the connec-tion between computer use and worker characteristics such as age,gender, and education. The most recent supplement includes asubstantial set of questions about the use of computers and theInternet at home, work, and school. The CPS and the BLS Occupa-tional Establishment Survey offer opportunities to assess how the

mix of occupations and, thus, skill types is changing in response tothe emerging digital economy. An open question is whether theoccupation codes need to be revised to reflect the changing natureof skills and tasks involved in the digital economy.

Price Behavior

Quality-adjusted deflators for computers have been in use for anumber of years, and this has greatly helped in quantifying theimpact of the IT revolution. Clearly this program must continue

Box 5

Data Needs for the Digital Economy: Price Behavior

Price deflators for goods and services must be adjusted to reflect changes inquality induced by IT. This will allow us to do more accurate measurementsof changes in key aggregate statistics such as productivity. Measures of pricedifferentials across goods and services sold by different methods (e.g., e-commerce vs. traditional methods) as well as measures of price dispersionacross producers using the same method are of critical importance tounderstanding the changing nature of competition in the digital economy.


32/408


since the performance of computers continues to increase while

their nominal prices continue to fall. Furthermore, as computertechnology becomes embedded in a growing number of otherproducts, we must ensure that the appropriate quality-adjusteddeflators are constructed for these as well.

Little thought or effort has been devoted to the impact of e-commerce on output price behavior. The ability of purchasers touse the Internet to search for the best price and other changes in

distribution channels that have the potential to eliminate whole-sale and retail markups may have important implications for boththe CPI and PPI programs.

What Can the Census Bureau and Other Statistical Agencies Doto Improve Our Understanding of the Digital Economy?

It is clear from the discussion so far that there are many holes in thedata collection efforts of the federal statistical system that needfilling before a clear understanding of the digital economy canemerge. There are many difficult and longstanding measurementand data collection issues that arise again in the context of measur-ing the digital economy. Important examples include defining andmeasuring output in the non-goods-producing sectors, collecting

establishment-level data from multiestablishment companies, andissues surrounding industry, commodity, and occupation classifica-tion systems. The digital economy has exacerbated many of theseproblems by spawning new products and services, new deliverymethods and forms of communication, and improved data-pro-cessing capabilities. The result is a rapidly changing businessenvironment that poses many challenges to agencies not known forrapid change. We are optimistic, however, that there are severalpractical and feasible steps that agencies can take to fill some ofthese data holes. Below are some examples.

Infrastructure

We should consider improving how we measure investment anddepreciation of IT and software. This would go beyond currentefforts with the ACES to break out equipment investment by typeof equipment. In particular, plant- (or some other subfirm-) level


33/408


measures are preferable if we are to assess the effects of these

investments on productivity, employment, and firm and industrystructure. (This is because many large firms span several industries,sectors, and geographic regions, and these firms account for a largeshare of investment in IT. Thus, it is not possible to get accuratemeasures of IT investment by industry or by geographic area withfirm-level surveys.) Some of this could be accomplished by aug-menting current data collection efforts. For example, questions on

IT investment could be added to the Economic Censuses. Annualplant-level data could be collected for manufacturing via the

Annual Survey of Manufactures. Outside of manufacturing, otherannual business surveys could be used to collect IT investment data.(The ASM is a plant-level survey. The annual surveys outside ofmanufacturing are establishment-based for single-unit firms. Inthe case of multiunit firms, however, these surveys typically use a

unit of observation based on business unitthat is, EI-line ofbusinessand, therefore, are not exactly plant- or firm-level sur-

veys.) While we should try to improve measures of the IT infrastruc-ture for all sectors of the economy, the manufacturing, services,

wholesale, and retail sectors should get the highest priority.Unfortunately, many large multiestablishment firms find it diffi-

cult to report investment and other items at the establishmentlevel. This is especially true outside of manufacturing. The CensusBureau and other statistical agencies need to work with businessesto get data at the lowest level of aggregation that firms can provide,so that agencies can provide the richest possible data for researchand policy analysis at a reasonable cost to the taxpayer.

E-Commerce

To get a handle on the extent and magnitude of e-commerce, wesuggest that the Census Bureau include class-of-customer andmethod-of-selling questions on all Economic Censuses and AnnualSurveys. These questions ask respondents to break out revenue bytype of buyer (e.g., consumers, businesses, government) and by

transaction method (e.g., in-store, mail order, Internet). Simplecross tabs could then provide estimates of business-to-business andbusiness-to-consumer e-commerce alongside traditional commerce.


34/408


Questions of this type are typically asked only in the retail, whole-

sale, and service sectors and are used primarily for classificationpurposes. The Internet and other direct-marketing channels haveincreased the need for such questions in the goods-producingsectors as well.

Classification efforts are particularly important for examining e-commerce. Under NAICS, businesses engaged primarily in Internetcommerce are classified separately from traditional retailers. This

is consistent with maintaining a production-oriented classifica-tion system. However, we still want to know how many books aresold. Thus, survey forms for Internet retailers should break outrevenues by commodity types. Currently, statistical agencies in theUnited States, Canada, and Mexico are developing a North Ameri-can Product Classification System (NAPCS) as the product classifi-cation companion to NAICS. This system should be designed with

e-commerce and the digital economy in mind.We expect that the impact of e-commerce on the markets for

digital goods (e.g., books, CDs, stock quotes) and services will bemuch larger than for goods and services that must be physicallydelivered (e.g., furniture, haircuts, pizza). Digital products arecharacterized by high fixed costs (e.g., writing a book) and lowmarginal costs (e.g., emailing a PDF file of a book; see Shapiroand Varian 1999). This has important implications for the opera-tion and structure of the markets for these goods and services, forintellectual property rights, for local tax authorities, and for inter-national trade (the Internet has no customs posts). Thus, it isimportant that we try to track the sales of digital goods and servicesby method of delivery. Currently, the limited bandwidth of theInternet limits this area of e-commerce., but improved technology

will allow for increased electronic delivery of such goods.Finally, we might consider undertaking an occasional survey that

examines e-commerce practices in the economy. This would in-clude asking firms how they use IT to communicate with suppliersand customers, whether they purchase or sell goods and serviceselectronically, and whether they use the Internet or other telecom-

munication networks for customer service and related tasks. The1999 ASM contained questions that address some of these issues. Ifthe Bureau is successful in collecting this information, it should


35/408


consider expanding inquiries of this type to surveys for other

sectors. This might also include surveying consumers on theirelectronic buying habits, perhaps through the Consumer Expendi-ture Survey. An important goal for such a consumer survey wouldbe to compare prices paid for similar goods and services purchasedelectronically and through traditional retail outlets.

Firm and Industry Structure

The Census Bureau and the Bureau of Labor Statistics already havemuch of what is required to examine the impact of investments inIT and the growth of e-commerce on the structure of firms andindustries. In particular, the Bureaus Standard Statistical Estab-lishment List has basic data on employment, payroll, industry, andlocation for the universe of employer business establishments in

the United States. The data can be linked to other Census Bureauestablishment-level and firm-level surveys. In this way, one couldcompare how the structure of IT-intensive firms changes over timerelative to less IT-intensive firms. An important question in thisarea is whether lower transaction costs associated with business-to-business e-commerce are leading to flatter firm organizationalstructures. For example, instead of relying of internal sources of

supply and support, firms that exploit e-commerce, with its associ-ated lower transaction costs, may now outsource these functions toother firms. If we combine data collected following our suggestionsabove with the SSEL, we expect to see firms that use e-commerceextensively shedding establishments that are outside the firmsmain line of business.

Another important issue is how the different marketing channels

made available by electronic networks are changing the structureof markets. Not only can firms set up an electronic storefront on theInternet and serve customers all over the world, but goods produc-ers can market directly to consumers and avoid traditional distribu-tion channels (e.g., manufacturer to wholesaler to retailer toconsumer). Thus, traditional boundaries defined by geographyand industry are being blurred. The SSEL linked to surveys askingabout class of customer and method of selling is the best way to seehow the structure of the economy is shifting from the traditionalmodel to the digital model.


36/408


Demographic and Worker Characteristics

We need to understand how both consumers and workers in thedigital economy differ from those in the traditional economy. TheConsumer Expenditure Survey should be modified to describe thedigital consumer better. First, household spending on computersand IT equipment and related expenditures (e.g., fees for Internetaccess) should be broken out separately. Next, the CES should ask

about the magnitude and nature of household e-commerce pur-chases (how much was spent, and on what goods and services). Ina similar vein, special supplements to the Current PopulationSurvey should continue to ask questions about computer andInternet use at home, school, and work. The precise nature of thesequestions should evolve so that they track the evolving role ofcomputers and the Internet in our activities.

Also, just as industry coding requires further consideration,occupation codes should be examined to determine whether theyneed to be modified to reflect the changing structure and tasks ofthe workforce. Modified occupation coding and related workforcecomposition change questions are relevant not only for householdsurveys but also for business surveys such as the BLS OccupationEstablishment Survey that measure and characterize changes inthe structure of the workforce.

Price Behavior

It will be important to quantify the impact the IT revolution and e-commerce are having on the prices businesses charge for goodsand services, many of which have been undergoing, and willcontinue to undergo, major quality changes. We are also interestedin whether e-commerce is changing price-cost margins and thenature of competition. For capturing quality change, we mustcollect information about the characteristics of goods and servicessold. Understanding changes in the nature of competition requirescollection of information about the pricing of goods sold over the

Internet and that of the same goods sold through more traditionalmethods. In this regard, it would be useful to quantify how price-cost markups have changed and how price dispersion across sellersof the same product varies by method of selling and, in the case of


37/408


digital products, by method of delivery to the consumer. Since

prices are traditionally collected by the BLS for the CPI and PPIprograms, coordination between BLS and Census about method ofselling and pricing behavior seems essential.

Other Areas

There are some more general ways in which we can modify the

federal statistical system to improve measurement of the digitaleconomy. First, we can improve our ability to measure output andproductivity in the non-goods-producing sectors. Second, we cancontinue to refine our industry, product, and input classificationsystems and increase the resources devoted to assigning establish-ments and businesses to appropriate categories. Third, we canincrease the resources devoted to developing and maintaining a

master list of business establishments, such as the SSEL, with high-quality industry, location, business age, and size information. This

would be an invaluable tool for providing a comprehensive per-spective on the changing landscape of business activity. Fourth, wecan increase the collection of micro-level data on businesses andhouseholds. Such data would allow us to control for relevant

worker and business characteristics and to compare businesses andworkers that have differentially adopted new processes and differ-entially produced new products and services. Moreover, as dis-cussed above, linking comprehensive files, such as the SSEL, tomicro data from specific targeted surveys allows us to shed light onhow changing business practices have influenced firm and industrystructure. The newly developed (and proposed) databases linkingemployer-employee data will also be valuable for examining theimpact that the digital economy is having on both businesses andthe workers within those businesses.

Discussion and Conclusions

While the ubiquity of IT is self-evident, our ability to quantify its

impact on the economy is limited by the nature and types of datacurrently being collected by federal statistical agencies and othersources. There are a number of unresolved conceptual questions


38/408


that exacerbate the measurement difficulties. For instance, the IT

revolution is closely connected to the growth of sectors of theeconomy (e.g., services) that we have traditionally struggled tomeasure.

The digital economy is forcing statistical agencies to rethink howthey measure the basic building blocks of our national accounts:outputs, inputs, and prices. Some progress has is being made onrefining the measurement of individual components (e.g., na-

tional investment in computers and the fraction of retail salesattributable to e-commerce). Clearly, policy and research needsrequire further efforts by statistical agencies to improve datacollection and measurement of the digital economy.

It is not likely that all the suggestions that we and others haveoffered can be implemented. We recognize that while policymakersand researchers have an insatiable appetite for data, concerns

about respondent burden and the resource costs of collecting datacannot be ignored. Realistic priorities must therefore be set by thedata-using community. We suggest that suggestions for changes tothe data collection programs at U.S. federal statistical agencies bemade within the following framework:

Plans to measure the digital economy should complement the

basic and longstanding programs of the U.S. statistical system thatmeasure the characteristics, inputs, outputs, and prices of busi-nesses and the characteristics and activities of individuals andhouseholds. The focus should be on measuring changes in thequality and use of IT and its impact on all sectors of the economy.There should be a special focus on improving measurement insectors such as services where measurement has traditionally been

difficult but there have been large investments in IT. Plans to measure the digital economy should leverage existingdata resources in a variety of ways including: development and useof underutilized administrative data sources, such as the SSEL;addition of supplementary questions to existing surveys and cen-suses; and encouragement of micro-level data development, in-cluding linking data from different sources and sharing data acrossdifferent U.S. federal statistical agencies.


39/408


In short, we suggest an incremental approach that modifies and

keeps intact our basic system for economic and demographicmeasurement.

In spite of this apparent caution, it is also important to recognizethat making changes in the basic data collection plans of the U.S.statistical agencies is a very slow process. For example, the newindustrial classification system, NAICS, is being implemented bythe statistical agencies over a 7-year horizon, and even though it is

a great advance over the prior system, it does not adequatelycapture the changes emerging from the growth of e-commerce.Moreover, plans are being made now for the next EconomicCensus in 2002. The inherently slow process of altering the courseof U.S. data collection activities implies that, unless we makeprogress in our thinking and plans now, we may find ourselves withrelatively little information about the magnitude, scope, and im-

pact of e-commerce for another decade or more.Put differently, U.S. statistical agencies need to set priorities now

in order to implement specific data collection plans. This paperintentionally stops short of setting these priorities. Instead, we havesought to provide a menu of measurement concerns and havestressed some general considerations that should be taken intoaccount in planning how to improve measurement of the digitaleconomy.

Acknowledgments

Opinions, findings, or conclusions expressed here are those of the authors anddo not necessarily reflect the views of the Census Bureau or the Department ofCommerce. We thank B. K. Atrostic, Erik Brynjolfsson, Frederick T. Knickerbocker,

and Thomas Mesenbourg for very helpful comments on earlier drafts of thispaper. This paper was written while John Haltiwanger served as Chief Economistof the Census Bureau.

References

Anonymous, 1999. Computer Age Gains Respect of Economists,New York Times,

April 19.Autor, D. H., L. F. Katz, and A. B. Krueger, 1997. Computing Inequality: HaveComputers Changed the Labor Market?NBER Working Paper No. 5956.


40/408


Berndt, E., and C. Morrison, 1995. High-tech Capital Formation and Economic

Performance in U.S. Manufacturing Industries: An Exploratory Analysis.Journalof Econometrics65: 943.

Brynolfsson, E., and L. Hitt, 1995. Computers as a Factor of Production: TheRole of Differences Among Firms.Economics of Innovation and New Technology3(May): 183199.

Brynolfsson, E., and L. Hitt, 1996. Paradox Lost? Firm-level Evidence on theReturns to Information Systems Spending.Management Science42(4): 541558.

Brynolfsson, E., and S. Yang, 1996. Information Technology and Productivity: AReview of the Literature.Advances in Computers43: 179214.

David, P., 1990. The Dynamo and the Computer: A Historical Perspective on theModern Productivity Paradox.American Economic Review Papers and Proceedings80(2): 355361

Dunne, T., L. Foster, J. Haltiwanger, and K. Troske, 1999. Wage and ProductivityDispersion in U.S. Manufacturing: The Role of Computer Investment.Mimeo,Center for Economic Studies, U.S. Bureau of the Census, Washington, DC.

Greenan, N., and J. Mairesse, 1996. Computers and Productivity in France:Some Evidence.NBER Working Paper No. 5836.

Greenwood, J., and M. Yorgulu, 1997. 1974.Carnegie-Rochester Conference Series onPublic Policy46: 4996.

Hitt, L., and E. Byrnolfsson, 1997. Information Technology and Internal FirmOrganization: An Exploratory Analysis.Journal of Management Information Systems14(2): 81101.

Jorgenson, D., and K. Stiroh, 1995. Computers and Growth.Economics ofInnovation and New Technology3 (May): 295316.

OECD, 1999. The Economic and Social Impacts of Electronic Commerce: PreliminaryFindings and Research Agenda. Paris: OECD.

Shapiro, C., and H. Varian, 1999. Information Rules: A Strategic Guide to the NetworkEconomy.Boston: Harvard Business School Press.

Siegel, D., and Z. Griliches, 1994. Purchased Services, Outsourcing, Computers,and Productivity in Manufacturing.In Z. Griliches et al., eds., Output Measure-ment in the Services Sectors, NBER Studies in Wealth, vol. 56 (Chicago: University ofChicago Press).

Solow, R., 1997. New York Times Book Review, July 12.

U.S. Department of Commerce, 1997. The Emerging Digital Economy. Washington,DC: Government Printing Office.


41/408

GDP and the Digital Economy: Keeping up with theChanges

Brent R. Moulton

The perception is widely held that the growth of the digitaleconomy1is unprecedented and has been a major contributor to

recent economic growth, the booming stock market, and therevival of productivity. What do we know about the growth of thedigital economy? What would we like to know that the data cur-rently do not reveal? And what does the federal statistical systemneed to do to provide that information? Because the economic datado not tell an unambiguous story about the digital economy,knowledgeable observers disagree about the importance of infor-

mation technology (IT) and electronic commerce in the economy.Economists have been engaged in a debate over the so-called

productivity paradox, which asks how productivity growth couldhave slowed during the 1970s and 1980s in the face of phenomenaltechnological improvements, price declines, and real growth incomputers and related IT equipment.2Much of this debate hasrevolved around questions of measurementfor example, are theoutput and growth of industries that use IT equipment beingadequately measured? There are reasons to think that they are not,that is, that the measures of output for the banking, insurance, andseveral other industries are particularly problematic, and the mea-sured productivity of these industries appears to be implausiblylow. If productivity in IT-using industries is not being measuredadequately, can the measurement errors explain the productivityparadox?3Several economists think that measurement may be animportant piece of the solution to the puzzle.


42/408

GDP and the Digital Economy35

In addition, the IT revolution has raised questions about the

ability of the federal statistical system to keep up with a changingeconomy. The availability of inexpensive IT equipment and ser-

vices has enabled businesses to do their work in new ways and hasled to the creation of new firms and even entire industries. Arethese new forms of business and production being adequatelycounted in our gross domestic product (GDP)? Have our economicstatistics kept up with electronic commerce, new kinds of financial

services, and new methods of inventory and product distribution?The economic data produced by the Department of Commerce

are critically valuable to our nations economic information infra-structure. The monthly releases of GDP are meticulously followedby policymakers and financial analysts, serving as a barometer ofthe economys health. These economic data provide informationfor understanding major policy issues, for forecasting the economys

potential for future growth, for conducting monetary policy, forunderstanding the tradeoffs between inflation and full employ-ment, for projecting tax revenues and conducting fiscal policy, andfor studying long-term issues such as the future of the social securitysystem. While these data serve as very good indicators of overalleconomic activity, they must constantly be improved and refined tokeep up with our rapidly evolving economy.

What Is Measured Well?

There are many aspects of IT and electronic commerce that aremeasured well in the official statistics. Some features of the digitaleconomy are captured perfectly well by the same data collectionsthat regularly provide information about the rest of the economy.The U.S. economic statistics for product and income arebenchmarked to input-output tables that are painstakingly con-structed from data collected in the economic censuses. The in-comes earned from production are benchmarked to tax andadministrative data. Adjustments are made to remove any sourcesof bias that are known and measurable. Because the IT and

electronic commerce sectors, like most other sectors, are coveredby the economic censuses, tax statistics, and unemployment insur-ance programs, data on the digital economy enter into the overall


43/408

36Moulton

measure of how the economy is doing in general. The GDP and

other basic economic statistics have been shown to provide verygood information about the movements over the business cycle ofproduction, aggregate demand and its components, income, andprices.4

Because the digital economy is not a standard classification foreconomic data, there may be some disagreement on what it entails.However it is defined, though, as a share of total GDP it is still fairly

small. (For example, private investment in information-processingequipment and software, a component of nonresidential fixedinvestment, was $407 billion in 1999, or 4.4 percent of GDP. At thispoint, Census Bureau estimates of the magnitude of electroniccommerce are more speculative but are still quite small as apercentage of all retail and wholesale sales.) Furthermore, at leastso far, movements in IT investment have not been highly correlated

with the ups and downs of the business cycle. Consequently, themeasurement problems that are central to the debate about theeffects of IT on long-term growth and productivity are not ques-tions about the usefulness of the national economic accounts formeasuring the short-term movements of the business cycle. Ratherthey are questions about small biases or omissions that amount toperhaps tenths of a percent per year, but that cumulatively affect

the measurement of long-term trends in growth and productivity.The Bureau of Economic Analysis (BEA) within the Department

of Commerce has tracked the direct effect of computers on mea-sured GDP growth using its contributions to percent changemethodology.5The contribution to the percent change of GDP canbe approximated by simply excluding the computer componentsin the various sectors of GDP (e.g., private fixed investment,

personal consumption expenditures, government gross invest-ment) in its calculation, and comparing the growth rate of realGDP less computers to the growth rate of real GDP. These data arenow regularly published in the GDP news release and are alsoavailable from the BEAs web site. As shown in table 1, the directcontribution of final sales of computers to real GDP growth aver-aged about 0.10.2 percentage point per year from 1987 to 1994,then accelerated to 0.30.4 percentage point per year from 1995 to1999. The acceleration reflected both increases in current-dollarfinal sales and more rapid declines in computer prices, and sug-


44/408

GDP and the Digital Economy37

gests that computers have recently become more important to thebusiness cycle.

The measurement of real growth of computers in the nationalaccounts is an example of a major statistical successan importantaspect of information technology that is now being more accuratelymeasured and better understood than it was a decade or two ago.

Fifteen years ago there was no adequate official price index forcomputers. Nearly everyone recognized that the price of comput-ing had been falling dramatically, but the methods used by the

Bureau of Labor Statistics (BLS) and the BEA for estimating priceindexes could not adequately account for quality changes of themagnitude that were occurring in computers.

The computer price problem was resolved through an excep-tional collaboration between a government agency (BEA) andindustry (in the form of a team of researchers from IBM). Theresearch group included people with technological and engineer-

ing knowledge as well as economists and statisticians. The quality-adjusted computer price index, which was introduced in thenational accounts in December 1985, helped rewrite economic

Table 1 Real GDP, Final Sales of Computers, and GDP Less Final Sales of Computers

GDP less final sales

GDP of computers Final sales of computers

(% change) (%change)1 Difference (% change)

1987 3.4 3.2 .2 23.4

1988 4.2 4.0 .2 20.3

1989 3.5 3.4 .1 13.4

1990 1.8 1.7 .1 5.6

1991 .5 .6 .1 12.0

1992 3.0 2.9 .1 24.8

1993 2.7 2.5 .2 22.1

1994 4.0 3.9 .1 20.1

1995 2.7 2.3 .4 53.7

1996 3.6 3.2 .4 55.3

1997 4.2 3.9 .3 45.4

1998 4.3 3.9 .4 53.9

1999 4.2 3.8 .4 44.1


45/408

38Moulton

history. The price index showed a remarkable multidecade decline

in prices and growth in output of computers and peripheralequipment.6Application of the new index resulted in significantlyhigher real economic growth. The method that was used to adjustfor quality improvements in the BEA computer price index has alsobeen adapted by the BLS for the computer components of itsproducer, consumer, export, and import price indexes.

Since 1985, the work on quality-adjusted price indexes has been

extended to several other IT products, such as semiconductors andtelephone switching equipment.7The BEA introduced improvedprice indexes for some types of software as part of the comprehen-sive revision of the national economic accounts released in fall1999. I must acknowledge, however, that progress on improvedmeasures of output and prices for high-tech products has been slowand difficult. Developing the statistical estimates that are required

for state-of-the-art quality adjustment is a resource-intensive activ-ity, and the necessary data and other resources have not alwaysbeen available.

Another success story in measuring the economic effects ofinformation technology was the elimination of substitution bias(that is, the tendency of indexes with fixed weights to overstategrowth). Prior to 1996, the national accounts measured changes inreal(that is, inflation-adjusted) product by holding prices con-stant at their levels during a particular base year. It was known thatthis method led to a distortion or bias as prices moved away fromthe levels of the base year, but it was generally assumed that changesin relative prices tended to be modest and that this bias couldtherefore be ignored. Once the improved price index for comput-ers was introduced, however, it became clear that its extreme andsustained downward trend wreaked havoc on the constant-pricemeasures of real GDP. The substitution bias caused the estimates ofreal GDP growth to be overstated by as much as a percentage point.Furthermore, because the bias was not constant over time, it led tosignificant distortions in measuring the long-term trends in growth.

The BEA embarked on a research program that eventually led to

the

brynjolfsson, erik & brian kahin (eds.) (2000). understanding the digital economy - data, tools, and...

Documents